Incremental processing of speech leads listeners to build expectations and to make predictions about upcoming sentence information. Prediction is, however, subject to individual variation (Huettig & Janse, 2016, doi:10.1080/23273798.2015.1047459) and may be compromised in L2 listeners (Kaan, 2014, doi:10.1075/lab.4.2.05kaa). This eye-tracking study explores the time course of L2 sentence processing and the (anticipatory) integration of different types of linguistic information in different types of noise. Using a visual world paradigm, we tested predictive processing among 72 German L2 listeners of English across three acoustic conditions (quiet, stationary noise, multi-talker babble) on three linguistic conditions (1-3).

(1) The girl is laugh-ING/laugh-ED at (by) the boy. (morpho-syntactic)

(2) The man is looking at the BIG/SMALL green ball. (lexical)

(3) The baby is looking at A/THE red purse. (discourse)

Noise can impact information integration in differential ways. Processing of morpho-syntactic cues, such as in (1), tends to be more affected by noise than lexical cues (2) and low-salience discourse cues, like articles (3). At the same time, L2 listeners generally tend to rely more on discourse than morphosyntactic cues (Cunnings, 2017, doi:10.1017/S1366728916001243). Hence, L2 listeners may differ in the types of information they continue to recover in noise as well as in their extent. In addition, different noise types mask speech differentially. Whereas stationary speech-shaped noise serves as an energetic masker, multi-talker babble functions as an informational masker. The informational masking effect is, however, modulated by language proficiency (Mattys et al., 2012, doi:10.1080/01690965.2012.705006). We investigate (a) whether L2 listeners continue to predict during sentence comprehension even when noisy speech causes phonetic unreliability, and we examine (b) the degree to which different linguistic cues are affected by the two types of noise in L2 listeners. Preliminary results suggest that both noise types equally affect the processing of lexical information, whereas morpho-syntactic processing is differentially affected by stationary vs. babble noise. The processing of discourse information, by contrast, is not differentially affected by the acoustic conditions. These results suggest that stimulus-specific and listener-specific aspects interact in predictive processing. We discuss potential implications for models of L2 language processing and models of speech in noise.